analytical framework on credit risks for ......asia-pacific development journal vol. 21, no. 2,...

Asia-Pacific Development Journal Vol. 21, No. 2, December 2014

1

ANALYTICAL FRAMEWORK ON CREDIT RISKS FORFINANCING SMALL AND MEDIUM-SIZED

ENTERPRISES IN ASIA

Naoyuki Yoshino and Farhad Taghizadeh-Hesary*

Small and medium-sized enterprises (SMEs) account for the major shareof employment and dominate the Asian economies. These economies areoften characterized as having bank-dominated financial systems andunderdeveloped capital markets, in particular venture capital. Hence,offering new methods for financing SMEs is crucial. Hometowninvestment trust funds are a form of financial intermediation that wasstarted recently and has since been adopted as a national strategy inJapan. In the present paper, the authors explain the importance of SMEsin Asia and describe hometown investment trust funds. They then providea scheme for credit rating of SMEs, employing two statistical analysistechniques, principal component analysis and cluster analysis to analysethe credit risks of a sample of Asian SMEs by using their financialvariables. This comprehensive and efficient method would enable banks,to group their SME customers based on their financial health, adjustinterest rates on loans and set lending ceilings for each group. Moreover,this method is applicable to hometown investment trust funds around theworld.

JEL Classification: G21, G23, G24, G32.

Key words: SME credit rating, SME financing, hometown investment trust fund.

I. INTRODUCTION

Small and medium-sized enterprises (SMEs) are the backbone of theeconomies in Asia. Over the period 2007-2012, they accounted for 98 per cent of all

* Naoyuki Yoshino, PhD, Dean, Asian Development Bank Institute (ADBI), Professor Emeritus, KeioUniversity, Japan (e-mail: [email protected]); and Farhad Taghizadeh-Hesary, PhD, Assistant Professorof Economics, Keio University, and Research Assistant to the Dean, Asian Development Bank Institute(ADBI), Japan (e-mail: [email protected], [email protected]).


2

enterprises and 38 per cent of the gross domestic product (GDP) on average andemployed 66 per cent of the national labour force (statistics in this paragraph fromADB, 2014). They also play a significant role in trade. Thirty per cent of total exportvalue was accounted for by SMEs in Asia on average during the above-cited period.In China, SMEs accounted for 41.5 per cent of total export value in 2012, up 6.8 percent year-on-year, while in Thailand, they accounted for 28.8 per cent of total exportvalue, growing 3.7 per cent year-on-year. SMEs that are part of global supply chainshave the potential to promote international trade and mobilize domestic demand.

Owing to the significance of SMEs to Asian national economies, it is importantto find ways to provide them with stable finance. Asian economies are oftencharacterized as having bank-dominated financial systems and capital markets thatare not well developed, particularly in the area of venture capital. Consequently, banksare the main source of financing. Although the soundness of the banking system hasimproved significantly since the 1996 Asian financial crisis, banks have been cautiousabout lending to SMEs, even though such enterprises account for a large share ofeconomic activity. Start-up companies, in particular, are finding it increasingly difficultto borrow money from banks because of strict Basel capital requirements. RiskierSMEs also face difficulty in borrowing money from banks (Yoshino, 2012). Hence, anefficient credit rating scheme that rates SMEs based on their financial health wouldhelp banks lend money to SMEs in a more rational way, while at the same time reducetheir risk.

Various credit-rating indices, such as Standard and Poor’s, rate largeenterprises. By looking at a large enterprise’s credit rating, banks can decide to lendthem up to a certain amount. However, for SMEs, the issue is more complicated asthere are no comparable ratings. The obstacles for setting up an SME credit ratingfacility are lack of data and difficulties in accessing the data to an authentic SMEdatabase. Nevertheless, there is a useful model in Japan. In a government-supportedproject, 52 credit guarantee corporations collected data from Japanese SMEs. Thesedata are stored at a private corporation called the Credit Risk Database, whichcontains data from 14.4 million SMEs, including default data from 3.3 millioncorporations and sole proprietors. If similar systems could be established in otherparts of Asia to accumulate and analyze credit risk data, and to accurately measurethe credit risk of each SME, then banks and other financial institutions could use it tocategorize their SME customers based on their financial health. SMEs would alsobenefit as they could raise funds from banks more easily and gain access to the debtmarket by securitizing their claims. Having a centralized SME database, such as theCredit Risk Database, is needed in other Asian countries and could be the long-runtarget for governments. In the short run, there are a number of available databasesthat can be used for credit rating. For example, the financial statements of SME


3

customers of governmental and private institutions could be used for credit rating ofbank customers or the extensive databases of SMEs held by tax bureaux of theministries of finance could be used as a database for the credit rating of SMEs.

In addition to banks, the creation of regional funds (or hometown investmenttrust funds) will help promote lending to start-up companies and riskier borrowers,such as SMEs. Selling these regional trust funds through branch offices of regionalbanks, post offices, credit associations and large banks will open up additionalsources for SMEs to raise funds.

In the present paper, section II contains a description of the importanceof SMEs and their difficulties in raising money. In section III, the advantages ofpreparing a complete SME database in each country is explained. This is followed bya discussion on an alternative way to provide stable financing for SMEs in Asia(hometown investment trust funds). In section IV, the authors propose a way ofestablishing SMEs’ credit ratings using statistical techniques and financial ratios. Thistakes into account the characteristics of SMEs, including leverage, liquidity,profitability, coverage and activity. The method can be used by banks around theworld to do the following: group SMEs based on their financial health; adjust interestrates on loans; and set lending ceilings for each group. Moreover, this method isapplicable for hometown investment trust funds. Section V contains concludingremarks.

II. IMPORTANCE OF SMALL AND MEDIUM-SIZED ENTERPRISESAND THEIR DIFFICULTIES TO RAISE FUNDS

Bank-dominated financial systems and the economic importance of small andmedium-sized enterprises in Asia

Figure 1 shows the size of the equity and bond markets and bank loans in Asia.

As indicated in figure 1, bank loans comprise the main share of the financialmarket in most Asian economies, and capital markets are not well developed in mostparts of the continent. This means that banks are the main source of SME financing.

Regarding the importance of SMEs in Asia, according to a survey conductedby the Asian Development Bank (ADB) (2014) on 14 countries from the five ADBregions: (a) Kazakhstan (Central Asia); (b) China and the Republic of Korea (East Asia);(c) Bangladesh, India, and Sri Lanka (South Asia); (d) Cambodia, Indonesia, Malaysia,the Philippines, Thailand, and Viet Nam (South-East Asia); and (e) Papua New Guineaand Solomon Islands (Pacific), SMEs, together with microenterprises, account formore than 90 per cent of total enterprises in each country.


4

Figure 1. Size of financial markets in Asia

Source: Shigesuke Kashiwagi, Nomura Holdings Inc., FSA Financial Research Center International Conference,Tokyo, Japan (February 2011).

Equity, bonds, and bank loans, as shares of total

Figure 2. Small and medium-sized enterprises contribution togross domestic product

Source: ADB (2014); Asia SME Finance Monitor 2013.

Note: Republic of Korea SME contribution to gross value added in manufacturing.

Kazakhstan

Republic of Korea

Indonesia

Malaysia

Thailand

0

10

20

30

40

50

60

70

2007 2008 2009 2010 2011 2012

(%)

%

100

90

80

70

60

50

40

30

20

10

0IndiaChina Malaysia Japan Republic of Korea

Equity Bonds Bank loans


5

SMEs, including microenterprises, contributed to 59.1 per cent of nominalgross domestic product (GDP) in Indonesia in 2012, a figure that is graduallyincreasing (figure 2). SMEs and microenterprises in Thailand contributed to 37.0 percent of nominal GDP in 2012, and in Malaysia, 32.7 per cent of real GDP in the sameyear. Thailand targeted an increase of the SME contribution to GDP to 40 per centor more in its country strategy of 2012. In Kazakhstan, the nominal GDP ofSMEs tended to increase but their contribution to GDP decreased over the period2010-2012, and was 17.3 per cent in 2012.

The extent of employment by SMEs varies by country (figure 3). The share ofSME employees to total employment ranged between 28.0 per cent (Kazakhstan) and97.2 per cent (Indonesia) in 2012.

Figure 3. Employment by small and medium-sized enterprises

KAZ

CHI

ROK

CAM

INO

MALPHI

THA

VIE

20

40

60

80

100

-5 0 5 10 15 20 25

SM

E e

mplo

yees to tota

l (%

)

SME employees growth (%)

Source: ADB, Asia SME Finance Monitor 2013 (Manila, 2014).

Notes: CAM = Cambodia, CHI = China, INO = Indonesia, KAZ = Kazakhstan,

ROK = Republic of Korea, MAL = Malaysia, PHI = Philippines, THA =Thailand, VIE = Viet Nam.

SME = small and medium-sized enterprise.

Data as of 2012 in China, Indonesia, Kazakhstan, Malaysia, Thailandand Viet Nam.

Data as of 2011 in Cambodia, Republic of Korea, and the Philippines.


6

Small and medium-sized enterprises face difficulties to raise funds

In comparison to large enterprises, SMEs find it more difficult to raise funds.This is because banks are reluctant to lend to them, even though these enterprisesaccount for a large share of the economic activity in their respective countries.

Figure 4 shows the results of survey on access to funding conducted by theBank of Japan. The two lines show how difficult or how easy it is to raise money fromthe markets. The thick line shows the difficulty faced by small enterprises and the thinline shows the same for the large enterprises. Data points below zero indicate thatcompanies are finding it difficult to raise money from banks or the capital markets.Small enterprises appear to be finding it more difficult to raise money in comparisonwith large firms.

Figure 4. Access to financing by small and medium-sized enterprisesand large firms in Japan

Source: Bank of Japan (2014).

Notes: DI = diffusion index; CY= calendar year.

30

20

10

0

-10

-20

-30

CY 95 97 99 01 03 05 07 09 11 13 14

Large enterprisesSmall enterprises


7

III. SMALL AND MEDIUM-SIZED ENTERPRISE DATABASEAND STABLE FINANCE

Considering the importance of SMEs to many dimensions of Asianeconomic activity, further efforts need to be made to offer them access to finance.Their financial and non-financial accounts are often difficult to assess. The Credit RiskDatabase in Japan, however, is being used to rate SMEs based on financial andnon-financial data. Extensive data on SMEs were collected for the Database and thenused to rate SMEs based on statistical analysis.

Database provided by the Credit Risk Database

The Credit Risk Database was founded in March 2001 as a membershiporganization to collect financial and non-financial data, including default information,on SMEs. It began as a voluntary association consisting of 52 credit guaranteecorporations in Japan. The Database was established with the objective to helpstreamline the process for obtaining SME financing and make it more efficient byassessing business conditions based on data and measuring credit risks related toSME financing.

As its membership and data collection expanded, the Credit Risk Databasebecame the source of data on SMEs. In April 2005, it obtained corporate status asa limited liability intermediate corporation and officially became the CRD Association.In June 2009, its status changed to a general incorporated association, as a result ofthe enforcement of the act on general incorporated associations and generalincorporated foundations. The CRD Association collects financial data on SMEs fromits members, namely credit guarantee corporations throughout Japan andgovernment-affiliated or private financial institutions involved in SME business. TheAssociation provides members with assessments of SMEs’ business situationsthrough a credit risk measurement model, which is based on the large amount ofcollected data (CRD website).1

The Credit Risk Database covers SMEs exclusively. As of March 2010, itincluded data on more than 50 per cent of the SMEs in Japan, covering more than14 million corporations and about 1.7 million sole proprietors, making it by far thelargest database for SMEs in Japan. The database for enterprises in default covered3,289,000 corporations and sole proprietors. Before the Credit Risk Database wasformally established, the Government of Japan invested 1.3 billion Japanese yen (¥)($12,000,000) from the supplementary budgets for fiscal years 1999 and 2000 to

1 www.crd-office.net.


8

finance the setting up of the Credit Risk Database computer system and otheroperational costs. The Association provides sample data and statistical information,as well as scoring services. Members of the CRD Association provide financial/non-financial data and default information on SMEs with whom members haverelationships (the names of SMEs are encoded so that they cannot be specified) tothe Database, which, in turn, returns to members a variety of services by utilizing theaccumulated data (figure 5).

Figure 5. Structure of Credit Risk Database

Source: Credit Risk Database of Japan website.

Establishing similar systems in other countries of Asia to accumulate andanalyze credit risk data and to measure credit risks of SMEs accurately would enablethem to raise funds from the banking sector and give them access to the debt marketby securitizing their claims.

Governments throughout Asia could set as a long-term target the developmentof a centralized SME database. In the interim, there are a variety of accessibledatabases that could be used for credit rating, such as the financial statements ofSME customers held by governmental and private institutions or the extensivedatabase of SMEs at the tax bureaux of the ministries of finance. In fact, data used forcredit analysis in section IV of the present paper is for a group of SMEs that arecustomers of an Iranian bank.

Members

(credit guarantee

corporations and

finance institutions)

Cleansing

Data consolidation

CRD Data CenterData on SMEs under the Small and MediumEnterprise Basic Law(1) Financial data(2) Non-financial data and attribute data(3) Default information

(2) (Provision of services)

(1) (Provision of data)

CRD services(1) Credit risk scoring(2) Data sampling(3) Statistical information

Member data

stored in

anonymous form

Database for use

(Accumulated CRD data)


9

Hometown investment trust funds

Given that the financial systems in Asia are dominated by banks, the creationof regional funds (or hometown investment trust funds)2 to promote lending tostart-up companies and riskier borrowers, such as SMEs, would help maintain thesoundness of the banking sector, as banks would not be exposed to the risks thatlending to such companies inevitably poses. Selling those regional trust funds throughbranch offices of regional banks, post offices, credit associations, and large bankswould increase funding sources for regional companies.

Such trust funds would not be guaranteed by a deposit insurance corporationand the associated risks would be borne by investors. The terms of a trust fund wouldhave to be fully explained to investors, such as where their funds would be investedand what the risks associated with the investment would be, in order to strengthenpotential investors’ confidence and help expand the trust fund market (Yoshino,2013). Examples of such funds in Japan include wind power generators andmusicians’ funds. In the first example, to construct 20 wind power generators,private-public partnerships were launched with investment of $1,000-5,000 by localresidents in a fund. They receive dividends every year through the sales of electricityby each wind power generator that they had invested in. Musicians’ funds gathermany small investors buying units for $150-500. If the musicians become successfuland their DVDs sell well, the sales generate a high rate of return for the fund.

Examples of both successful and failed funds can be cited. Project assessorsplay a key role in evaluating each project to limit the number of non-performinginvestments and losses by investors. Some of the funds set up in Japan are regardedas charities, with some investors viewing them as a way to invest in their region tosupport new business ventures.

Such new ventures pose a problem for banks, as although some will havehigh expected rates of return, the high risks involved make it difficult for banks tofinance them. However, if the projects are financed by hometown investment trustfunds rather than by deposits transformed into bank loans, they will not createnon-performing loans for banks. Banks can still benefit and compete with each otherby selling the trust funds through their branch offices, although it has to be madeclear that an investment in those funds is not guaranteed. If a bank sells successfulhometown investment trust funds, it will be able to attract more investors while on theother hand, if it sells loss-making funds, it will lose investors in the future. Competition

2 Hometown investment trust funds were only recently established and now have been adopted asa national strategy in Japan (Yoshino and Taghizadeh-Hesary, 2014a).


10

Depositors

Investors

Hometown

Investment

Trust Funds

(HIT)

Banks

Riskier

Borrowers

Safer

SMEs

will improve the quality of projects and enhance the risk-adjusted returns forinvestors.

Figure 6 shows how trust funds can increase investment in riskier projects.

Figure 6. Bank-based small and medium-sized enterprise financing andhometown investment financing to riskier borrowers

Source: Yoshino and Taghizadeh-Hesary (2015).

A hometown investment trust fund has three main advantages. First, itcontributes to financial market stability by lowering information asymmetry. Individualhouseholds and firms have direct access to information about the borrowing firms,mainly SMEs. Second, it is a stable source of risk capital. The fund is project driven.Firms and households decide to invest by getting to know the borrowers and theirprojects. In this way, the fund distributes risk, but not so that it renders riskintractable, which has been the problem with the “originate and distribute” model.Third, it contributes to economic recovery by connecting firms and households withSMEs that are worthy of their support. It also creates employment opportunities at theSMEs as well as for the pool of retirees from financial institutions who can help assessthe projects (Yoshino, 2013; Yoshino and Taghizadeh-Hesary, 2014b; 2014c).

IV. ANALYSIS OF SMALL AND MEDIUM-SIZED ENTERPRISECREDIT RISK USING ASIAN DATA

In this section an efficient and comprehensive scheme for credit rating ofSMEs is presented. Various financial ratios that describe the characteristics of SMEsand enable banks to categorize their SME customers into different groups based ontheir financial health are examined. This method is also applicable for hometown


11

investment trust funds. The data for this statistical analysis were provided by anIranian bank for 1,363 SMEs.

This method could be also used for credit rating even in the non-SME sector.For a recent study, Yoshino, Taghizadeh-Hesary and Nili (2015) used this method forcredit rating and classifying 32 Iranian banks. Based on the results, the banks wereclassified into two groups and rated based on their soundness.

Selection of the variables

A number of possible ratios have been identified as useful in predicting a firm’slikelihood of default. Chen and Shimerda (1981) show that out of more than 100financial ratios, almost 50 per cent of them were useful in at least one empirical study.Some scholars have argued that quantitative variables are not sufficient to predictSME default and that including qualitative variables, such as the legal form of thebusiness, the region where the main business is carried out and the industry type,improves the models’ predictive power (for example, Lehmann, 2003; Grunert, Nordenand Weber, 2004). However, the data used here are based on firm’s financialstatements, which do not contain such qualitative variables.

For this study, the author’s followed Altman and Sabato (2007), who proposedfive categories to describe a company’s financial profile: liquidity; profitability;leverage; coverage; and activity. For each of those categories, they created a numberof financial ratios identified in the literature. Table 1 shows the financial ratios selectedfor this survey.

In the next stage, two statistical techniques are used: principal componentanalysis; and cluster analysis. The underlying logic of both techniques is dimensionreduction, namely summarizing information on multiple variables into just a fewvariables, which is achieved in different ways. Principal component analysis reducesthe number of variables into components (or factors). Cluster analysis reduces thenumber of SMEs by placing them in small clusters. In this survey, components(factors) resulting from the principal component analysis are applied and then thecluster analysis is carried out in order to group the SMEs.

Principal component analysis

Principal component analysis is a standard data reduction technique thatentails extracting data, removing redundant information, highlighting hidden featuresand visualizing the main relationships that exist between observations.3 This analysis

3 Principal component analysis is also referred to as the Karhunen–Loève Transform (named after KariKarhunen and Michel Loève.)


12

is a technique for simplifying a data set by reducing multidimensional data sets tolower dimensions for analysis. Unlike other linear transform methods, principalcomponent analysis does not have a fixed set of basis vectors. Its basis vectorsdepend on the data set. The principal component analysis has the additionaladvantage of indicating what is similar and different about the various models created(Bruce-Ho and Dash-Wu, 2009). Through this method, the 11 variables listed intable 1 were used to determine the minimum number of components that can accountfor the correlated variance among SMEs.

Table 1. Examined variable

No. Symbol Definition Category

1 Equity_TL Equity (book value)/total liabilitiesLeverage

2 TL_Tassets Total liabilities/total assets

3 Cash_Tassets Cash/total assets

4 WoC_Tassets Working capital/total assets Liquidity

5 Cash_Sales Cash/net sales

6 EBIT_Sales EBIT/sales

7 Rinc_Tassets Retained earnings/total assets Profitability

8 Ninc_Sales Net income/sales

9 EBIT_IE EBIT/interest expenses Coverage

10 AP_Sales Account payable/salesActivity

11 AR_TL Account receivable/total liabilities

Notes: Retained earnings = the percentage of net earnings not paid out as dividends, but retained by thecompany to be reinvested in its core business or to pay debt. It is recorded under shareholders’ equity inthe balance sheet. EBIT = earnings before interest and taxes. Account payable = an accounting entrythat represents an entity’s obligation to pay off a short-term debt to its creditors. The accounts payableentry is found on a balance sheet under current liabilities. Account receivable = money owed bycustomers (individuals or corporations) to another entity in exchange for goods or services that havebeen delivered or used, but not yet paid for. Receivables usually come in the form of operating lines ofcredit and are usually due within a relatively short time period, ranging from a few days to a year.

In order to examine the suitability of these data for factor analysis, the Kaiser-Meyer-Olkin (KMO) test and Bartlett’s test of sphericity were performed. KMO isa measure of sampling adequacy that indicates the proportion of common variancethat might be caused by underlying factors. High KMO values (larger than 0.60)generally indicate that factor analysis may be useful, which is the case in this study:KMO = 0.71. If the KMO value is less than 0.5, factor analysis would not be useful.Bartlett’s test of sphericity indicates whether the correlation matrix is an identity


13

matrix, indicating that variables are unrelated. A significance level less than 0.05indicates that there are significant relationships among the variables, which is thecase in this study: significance of Bartlett’s test < 0.001.

Next, the number of factors to be used in the analysis was determined. Table 2reports the estimated factors and their eigenvalues. Only those factors accountingfor more than 10 per cent of the variance (eigenvalues > 1) were kept in the analysis.As a result, only the first four factors were finally retained (table 2).

Taken together, Z1 through Z4 explain 71.06 per cent of the total variance ofthe financial ratios.

Table 2. Total variance explained

Component Eigenvalues % of variance Cumulative variance %

Z1 3.30 30.00 30.00

Z2 2.19 19.90 49.90

Z3 1.25 11.38 61.28

Z4 1.08 9.78 71.06

Z5 0.94 8.56 79.62

Z6 0.75 6.79 86.41

Z7 0.56 5.09 91.50

Z8 0.48 4.36 95.86

Z9 0.32 2.87 98.73

Z10 0.13 1.14 99.87

Z11 0.09 0.13 100.00

To run the principal component analysis, the direct oblimin rotation was used.Direct oblimin is the standard method to obtain a non-orthogonal (oblique) solution —that is, one in which the factors are allowed to be correlated. In order to interpret therevealed principal component analysis information, the pattern matrix must then bestudied. Table 3 presents the pattern matrix of factor loadings by use of the directoblimin rotation method in which variables with large loadings, absolute value (> 0.5)for a given factor, are highlighted in bold.


14

As evident in table 3, the first component, Z1, has four variables with anabsolute value (> 0.5), of which two are positive (EBIT/sales and net income/sales)and two are negative (cash/net sales and account payable/sales). For Z1, thevariables with large loadings are mainly net income and earnings, hence, Z1 generallyreflects the net income of an SME. As this factor explains the most variance in thedata, it is the most informative indicator of the overall financial health of an SME. Z2reflects short-term assets. This component has three major loading variables:(a) liabilities/total assets which is negative and means an SME has few liabilities andmainly relies on its own assets; (b) working capital/total assets which is positive and,means an SME has short-term assets; (c) retained earnings/total assets, which ispositive and means an SME has earnings, which it reinvests in the company. Thesethree variables indicate an SME that has small borrowing and sizeable working capitaland retained earnings, and therefore, has plenty of short-term assets. Z3 reflects theliquidity of SMEs. This factor has two variables with large loadings (cash/total assetsand EBIT/interest expenses), both with positive values, which shows an SME that iscash-rich and has high earnings, hence, it mainly reflects the liquidity conditions of anSME. The last factor, Z4, reflects capital. It has two variables with large loading, bothwith positive values, equity (book value)/total liabilities and account receivable/totalliabilities and indicates an SME with few liabilities that has sizeable equity.

Table 3. Factor loadings of financial variables after direct oblimin rotation

Variables Component

(financial ratios) Z1 Z2 Z3 Z4

Equity_TL 0.009 0.068 0.113 0.705

TL_Tassets -0.032 -0.878 0.069 -0.034

Cash_Tassets -0.034 -0.061 0.811 0.098

WoC_Tassets -0.05 0.762 0.044 0.179

Cash_Sales -0.937 0.021 0.083 0.009

EBIT_Sales 0.962 0.008 0.024 -0.004

Rinc_Tassets 0.014 0.877 0.015 -0.178

Ninc_Sales 0.971 -0.012 0.015 0.014

EBIT_IE 0.035 0.045 0.766 -0.098

AP_Sales -0.731 -0.017 -0.037 -0.016

AR_TL 0.009 -0.041 -0.104 0.725

Note: The extraction method was principal component analysis. The rotation method was direct oblimin withKaiser normalization.


15

Table 4 shows the correlation matrix of the components. It indicates that thereis no correlation between these four components. This means a regular orthogonalrotation approach could have been used to force an orthogonal rotation. In thissurvey, an oblique rotation method is used. This method still provides basically anorthogonal rotation factor solution because these four components are not correlatedwith each other and are distinct entities.

Table 4. Component correlation matrix

Component Z1 Z2 Z3 Z4

Z1 1.00 0.037 -0.031 -0.005

Z2 0.037 1.00 0.106 0.102

Z3 -0.031 0.106 1.00 0.033

Z4 -0.005 0.102 0.033 1.00

Note: The extraction method is principal component analysis. The rotation method is directoblimin with Kaiser Normalization.

Figure 7 shows the distribution of the four components (Z1, Z2, Z3, and Z4) forgroup A, financially sound SMEs, and group B, non-sound SMEs.

It is clear from the six graphs in this figure that group A (sound) SMEs cangenerally be found in the positive areas of the graphs and group B SMEs in thenegative areas, or in inferior places in most cases when compared to the soundgroup. This shows that these four defined components (Z1, Z2, Z3, and Z4) are ableto separate SMEs. It means that these components could be a good measure forshowing the financial healthiness of SMEs.

Cluster analysis

In this section, the four components that were obtained in the previous sectionare reviewed to identify SMEs that have similar traits. Clusters were then generatedand SMEs were placed in distinct groups. To do this, cluster analysis is undertaken. Incluster analysis, a set of data is organized into groups so that observations froma group with similar characteristics can be compared with those from a differentgroup (Martinez and Martinez, 2005). In this case, SMEs were organized into distinctgroups according to the four components derived from the principal componentanalysis used in the previous section. Cluster analysis techniques can themselvesbe broadly grouped into three classes: hierarchical clustering, optimization


16

Figure 7. Distribution of factors for groups A and B of small andmedium-sized enterprises

Notes: Group A = sound SMEs, group B = non-sound SMEs. The firms considered to be non-sound in this studyhave risk-weighted assets greater than their shareholders’ equity.


17

clustering,4 and model-based clustering. For this study, the most prevalent method inthe literature, hierarchical clustering was used. This produces a nested sequence ofpartitions by merging (or dividing) clusters. At each stage of the sequence, a newpartition is optimally merged (or divided) from the previous partition according tosome adequacy criterion. The sequence of partitions ranges from a single clustercontaining all the individuals to a number of clusters (n) containing a single individual.The series can be described by a tree display called the dendrogram (figure 8).Agglomerative hierarchical clustering proceeds by a series of successive fusions ofthe n objects into groups. By contrast, divisive hierarchical methods divide the nindividuals into progressively finer groups. Divisive methods are not commonly usedbecause of the computational problems they pose, see Everitt, Landau and Leese(2001) and Landau and Chis Ster (2010). Below, the average linkage method, which isa hierarchical clustering technique, is used.

The average linkage method

The average linkage (AL) method defines the distance between clusters as theaverage distance from all observations in one cluster to all points in another cluster. Inother words, it is the average distance between pairs of observations, where one isfrom one cluster and one is from the other. The average linkage method is relativelyrobust and also takes the cluster structure into account (Martinez and Martinez, 2005;Feger and Asafu-Adjaye, 2014). The basic algorithm for the AL method can besummarized in the following manner:

• N observations start out as N separate groups. The distance matrix D =(dij) is searched to find the closest observations, for example, Y and Z.

• The two closest observations are merged into one group to form a cluster(YZ), producing N – 1 total groups. This process continues until all theobservations are merged into one large group.

Figure 8 shows the dendrogram that results from this hierarchical clustering.

4 The main difference between the hierarchical and optimization techniques is that in hierarchicalclustering, the number of clusters is not known beforehand. The process consists of a sequence of stepsduring which two groups are either merged (agglomerative) or divided (divisive) according to the level ofsimilarity. Eventually, each cluster can be subsumed as a member of a larger cluster at a higher level ofsimilarity. The hierarchical merging process is repeated until all subgroups are fused into a single cluster(Martinez and Martinez, 2005). Optimization methods on the other hand do not necessarily formhierarchical classifications of the data as they produce a partition of the data into a specified orpredetermined number of groups by either minimizing or maximizing some numerical criterion (Feger andAsafu-Adjaye, 2014).


18

The resultant dendrogram (hierarchical average linkage cluster tree) providesa basis for determining the number of clusters by sight. In the dendrograms shown infigure 8, the horizontal axis shows 1,363 SMEs. Owing to their large number, SMEshave not been identified by number in the dendrogram, although this is how SMEs areidentified in this survey. Rather, the dendrogram categorizes the SMEs in three mainclusters (groups 1, 2, and 3), but it does not show which of those three clusterscontain the financially healthy SMEs, which contains least healthy or risky SMEs, andwhich contain intermediate SMEs, hence, there is one more step to go.

Figure 8 shows the 1,363 SMEs categorized into three major clusters. Usingtheir components, which were derived from the principal component analysis, thedistribution of factors for each member of the three major clusters can be plotted.Figure 9 shows the distribution of Z1-Z2 for these three cluster members separately.5

5 The dendrogram shows us the major and minor clusters. One of useful features of this tree is that itidentifies a representative SME of most of the minor groups, which has the average traits of the othermembers of the group. Hence for simplification, in figure 9, there is only used data from theserepresentative SMEs, which explains the whole group’s traits. This is why the total number ofobservations in figure 9 is lower than the 1,363 observations in this survey.

Figure 8. Dendrogram using average linkage


19

As indicated in figure 9, group 1 consists of the healthest SMEs, group 3contains the SMEs with the lowest healthiness and group 2 are those that arein-between. Interestingly, when this grouping is carried out by using the othercomponents (Z1-Z3, Z1-Z4, Z2-Z4, Z2-Z3, and Z3-Z4), in most cases, the grouping issimilar. This implies that this analysis is an effective way of grouping SMEs.

V. CONCLUDING REMARKS

SMEs play a significant role in Asian economies as they are responsible forvery high shares of employment and output in all Asian countries. However, they havelimited access to finance compared to large enterprises. Banks dominate the financialsystems in Asia and as a consequence are the main source of financing for SMEs.Besides banks, the creation of regional funds (or hometown investment trust funds) topromote lending to riskier customers, such as SMEs, would be the second choice offinancing.

Figure 9. Grouping based on principal component analysis (Z1-Z2)and cluster analysis

Note: Group 1 = healthiest SMEs; group 2 = in-between SMEs; group 3 = least healthy SMEs.


20

For financial institutions, it is crucial to recognize healthy SMEs from thenon-healthy SMEs, in order to avoid accumulation of non-performing loans. This ispossible by applying statistical analysis techniques on financial variables of SMEs.

In this research, 11 financial variables of 1,363 SMEs that are customers of anAsian bank are used for the analysis. These variables are subjected to principalcomponent analysis and cluster analysis. The results showed that four variables (netincome, short-term assets, liquidity and capital) are the most important for describingthe general characteristics of SMEs. Three groups of SMEs were then differentiatedbased on the financial health of those 1,363 SMEs.

The policy implications of this paper are that it is crucial for the governments tocollect SME data and prepare a rich database, such as the Credit Risk Database ofJapan. This will help governments formulate economic policies and facilitate a pathtowards developing an efficient credit rating mechanism. Since establishing a creditrisk database takes time, in the short-run, governmental or private financialinstitutions could apply similar credit rating techniques on SME customers’ balancesheet information, and recognize healthy SMEs. Under such a scenario: (a) financiallyhealthy SMEs could borrow much more money from banks at lower interest ratesbecause of their lower default risk; (b) SMEs in poor financial health would have topay higher interest rates on their borrowing from the financial institutions and havea lower borrowing ceiling than healthy SMEs; and (c) banks could reduce the amountof their accumulated non-performing loans to SMEs. Also of note, if the hometowninvestment trust funds were to be sold through regional banks, post offices, creditassociations, or even large banks, these financial institutions could apply theaforementioned credit rating analysis, and decide on the closure or continuing of suchfunds.


21

REFERENCES

Altman, E.I., and G. Sabato (2007). Modelling credit risk for SMEs: evidence from the U.S. market.ABACUS, vol. 43, No. 3, pp. 332-357.

Asian Development Bank (ADB) (2014). Asia SME Finance Monitor 2013. Manila.

Bank of Japan (2014). Tankan 2014: short-term economic survey of enterprises in Japan. Tokyo:Research and Statistics Department.

Bruce-Ho, C.-T., and D. Dash-Wu (2009). Online banking performance evaluation using dataenvelopment analysis and principal component analysis. Computers & OperationsResearch, vol. 36, pp. 1835-1842.

Chen, K.H., and T.A. Shimerda (1981). An empirical analysis of useful financial ratios. FinancialManagement, vol. 10, No. 1, pp. 51-60.

Everitt, B.S., S. Landau, and M. Leese (2001). Cluster Analysis, 4 ed. London: Arnold.

Feger, T., and J. Asafu-Adjaye (2014). Tax effort performance in Sub-Sahara Africa and the role ofcolonialism. Economic Modelling, vol. 38, pp. 163-174.

Grunert, J., L. Norden, and M. Weber (2004). The role of non-financial factors in internal credit ratings.Journal of Banking and Finance, vol. 29, No. 2, pp. 509-531.

Landau, S., and I. Chis Ster (2010). Cluster analysis: overview. International Encyclopedia ofEducation, 3 ed. Oxford: Elsevier.

Lehmann, B. (2003). Is it worth the while? The relevance of qualitative information in credit rating.Working paper presented at the European Financial Management Association 2003 AnnualMeetings. Helsinki, 25-28 June.

Martinez, W.L., and A.R. Martinez (2005). Exploratory Data Analysis with Matlab. Florida: Chapmanand Hall/CRC Press.

Yoshino, N. (2012). Global imbalances and the development of capital flows among Asian countries.OECD Journal: Financial Market Trends, vol. 2012, No. 1, pp. 81-112.

(2013). The background of hometown investment trust funds. In Hometown InvestmentTrust Funds: A Stable Way to Supply Risk Capital, N. Yoshino and S. Kaji, eds. Tokyo:Springer.

Yoshino, N., and F. Taghizadeh-Hesary (2014a). Hometown investment trust funds: an analysis ofcredit risk. Paper presented at the Executive Secretary’s Seminar Series on Financing forSustainable Development: Local and Global Perspectives. United Nations Economic andSocial Commission for Asia and the Pacific. Bangkok, 31 October.

(2014b). Hometown investment trust funds: an analysis of credit risk. Working Paper,No. 505. Tokyo: Asian Development Bank Institute.

(2014c). Three arrows of “Abenomics” and the structural reform of Japan: inflation targetingpolicy of the Central Bank, fiscal consolidation, and growth strategy. Working Paper,No. 492. Tokyo: Asian Development Bank Institute.

(2015). An analysis of challenges faced by Japan’s economy and Abenomics. JapanesePolitical Economy Journal, vol. 40, pp. 1-26, DOI: 10.1080/2329194X.2014.998591.

Yoshino, N., F. Taghizadeh Hesary, and F. Nili (2015). Estimating dual deposit insurance premium ratesand forecasting non-performing loans: two new models. Working Paper, No. 510. Tokyo:Asian Development Bank Institute.

analytical framework on credit risks for ......asia-pacific development journal vol. 21, no. 2,...

Documents